blob: ca702c67bc6635b95ac2345811b65301c9ed0e57 [file] [log] [blame]
Sage Weilde576062009-10-06 11:31:07 -07001#ifndef _FS_CEPH_SUPER_H
2#define _FS_CEPH_SUPER_H
3
4#include "ceph_debug.h"
5
6#include <asm/unaligned.h>
7#include <linux/backing-dev.h>
8#include <linux/completion.h>
9#include <linux/exportfs.h>
10#include <linux/fs.h>
11#include <linux/mempool.h>
12#include <linux/pagemap.h>
13#include <linux/wait.h>
Stephen Rothwellf1a3d572010-01-18 11:53:08 +110014#include <linux/writeback.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090015#include <linux/slab.h>
Sage Weilde576062009-10-06 11:31:07 -070016
17#include "types.h"
18#include "messenger.h"
19#include "msgpool.h"
20#include "mon_client.h"
21#include "mds_client.h"
22#include "osd_client.h"
23#include "ceph_fs.h"
24
25/* f_type in struct statfs */
26#define CEPH_SUPER_MAGIC 0x00c36400
27
28/* large granularity for statfs utilization stats to facilitate
29 * large volume sizes on 32-bit machines. */
30#define CEPH_BLOCK_SHIFT 20 /* 1 MB */
31#define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
32
33/*
34 * mount options
35 */
36#define CEPH_OPT_FSID (1<<0)
37#define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */
38#define CEPH_OPT_MYIP (1<<2) /* specified my ip */
39#define CEPH_OPT_DIRSTAT (1<<4) /* funky `cat dirname` for stats */
40#define CEPH_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
41#define CEPH_OPT_NOCRC (1<<6) /* no data crc on writes */
42#define CEPH_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
43
44#define CEPH_OPT_DEFAULT (CEPH_OPT_RBYTES)
45
46#define ceph_set_opt(client, opt) \
Sage Weil6b805182009-10-27 11:50:50 -070047 (client)->mount_args->flags |= CEPH_OPT_##opt;
Sage Weilde576062009-10-06 11:31:07 -070048#define ceph_test_opt(client, opt) \
Sage Weil6b805182009-10-27 11:50:50 -070049 (!!((client)->mount_args->flags & CEPH_OPT_##opt))
Sage Weilde576062009-10-06 11:31:07 -070050
51
Sage Weilde576062009-10-06 11:31:07 -070052struct ceph_mount_args {
53 int sb_flags;
Sage Weil6b805182009-10-27 11:50:50 -070054 int num_mon;
55 struct ceph_entity_addr *mon_addr;
Sage Weilde576062009-10-06 11:31:07 -070056 int flags;
57 int mount_timeout;
Yehuda Sadehf5a20412010-02-03 11:00:26 -080058 int osd_idle_ttl;
Sage Weilde576062009-10-06 11:31:07 -070059 int caps_wanted_delay_min, caps_wanted_delay_max;
60 struct ceph_fsid fsid;
61 struct ceph_entity_addr my_addr;
62 int wsize;
63 int rsize; /* max readahead */
64 int max_readdir; /* max readdir size */
Yehuda Sadeh2baba252009-12-18 13:51:57 -080065 int congestion_kb; /* max readdir size */
Sage Weilde576062009-10-06 11:31:07 -070066 int osd_timeout;
Yehuda Sadeh422d2cb2010-02-26 15:32:31 -080067 int osd_keepalive_timeout;
Sage Weilde576062009-10-06 11:31:07 -070068 char *snapdir_name; /* default ".snap" */
Sage Weil4e7a5dc2009-11-18 16:19:57 -080069 char *name;
Sage Weilde576062009-10-06 11:31:07 -070070 char *secret;
71 int cap_release_safety;
72};
73
74/*
75 * defaults
76 */
77#define CEPH_MOUNT_TIMEOUT_DEFAULT 60
Yehuda Sadeh422d2cb2010-02-26 15:32:31 -080078#define CEPH_OSD_TIMEOUT_DEFAULT 60 /* seconds */
79#define CEPH_OSD_KEEPALIVE_DEFAULT 5
Yehuda Sadehf5a20412010-02-03 11:00:26 -080080#define CEPH_OSD_IDLE_TTL_DEFAULT 60
Sage Weil8fa97652009-10-16 14:44:35 -070081#define CEPH_MOUNT_RSIZE_DEFAULT (512*1024) /* readahead */
Sage Weilde576062009-10-06 11:31:07 -070082
83#define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024)
84#define CEPH_MSG_MAX_DATA_LEN (16*1024*1024)
85
86#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weil4e7a5dc2009-11-18 16:19:57 -080087#define CEPH_AUTH_NAME_DEFAULT "guest"
Sage Weilde576062009-10-06 11:31:07 -070088
89/*
90 * Delay telling the MDS we no longer want caps, in case we reopen
91 * the file. Delay a minimum amount of time, even if we send a cap
92 * message for some other reason. Otherwise, take the oppotunity to
93 * update the mds to avoid sending another message later.
94 */
95#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
96#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
97
98
99/* mount state */
100enum {
101 CEPH_MOUNT_MOUNTING,
102 CEPH_MOUNT_MOUNTED,
103 CEPH_MOUNT_UNMOUNTING,
104 CEPH_MOUNT_UNMOUNTED,
105 CEPH_MOUNT_SHUTDOWN,
106};
107
108/*
109 * subtract jiffies
110 */
111static inline unsigned long time_sub(unsigned long a, unsigned long b)
112{
113 BUG_ON(time_after(b, a));
114 return (long)a - (long)b;
115}
116
117/*
118 * per-filesystem client state
119 *
120 * possibly shared by multiple mount points, if they are
121 * mounting the same ceph filesystem/cluster.
122 */
123struct ceph_client {
Sage Weil07433042009-11-18 16:50:41 -0800124 struct ceph_fsid fsid;
125 bool have_fsid;
Sage Weilde576062009-10-06 11:31:07 -0700126
127 struct mutex mount_mutex; /* serialize mount attempts */
Sage Weil6b805182009-10-27 11:50:50 -0700128 struct ceph_mount_args *mount_args;
Sage Weilde576062009-10-06 11:31:07 -0700129
130 struct super_block *sb;
131
132 unsigned long mount_state;
Sage Weil9bd2e6f2010-02-02 16:21:06 -0800133 wait_queue_head_t auth_wq;
Sage Weilde576062009-10-06 11:31:07 -0700134
Sage Weil9bd2e6f2010-02-02 16:21:06 -0800135 int auth_err;
Sage Weilde576062009-10-06 11:31:07 -0700136
Sage Weil85ccce42010-02-17 10:02:43 -0800137 int min_caps; /* min caps i added */
138
Sage Weilde576062009-10-06 11:31:07 -0700139 struct ceph_messenger *msgr; /* messenger instance */
140 struct ceph_mon_client monc;
141 struct ceph_mds_client mdsc;
142 struct ceph_osd_client osdc;
143
144 /* writeback */
145 mempool_t *wb_pagevec_pool;
146 struct workqueue_struct *wb_wq;
147 struct workqueue_struct *pg_inv_wq;
148 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800149 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -0700150
151 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -0800152
153#ifdef CONFIG_DEBUG_FS
154 struct dentry *debugfs_monmap;
155 struct dentry *debugfs_mdsmap, *debugfs_osdmap;
156 struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800157 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -0800158 struct dentry *debugfs_bdi;
Sage Weil07433042009-11-18 16:50:41 -0800159#endif
Sage Weilde576062009-10-06 11:31:07 -0700160};
161
162static inline struct ceph_client *ceph_client(struct super_block *sb)
163{
164 return sb->s_fs_info;
165}
166
167
168/*
169 * File i/o capability. This tracks shared state with the metadata
170 * server that allows us to cache or writeback attributes or to read
171 * and write data. For any given inode, we should have one or more
172 * capabilities, one issued by each metadata server, and our
173 * cumulative access is the OR of all issued capabilities.
174 *
175 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
176 * session capability lists.
177 */
178struct ceph_cap {
179 struct ceph_inode_info *ci;
180 struct rb_node ci_node; /* per-ci cap tree */
181 struct ceph_mds_session *session;
182 struct list_head session_caps; /* per-session caplist */
183 int mds;
184 u64 cap_id; /* unique cap id (mds provided) */
185 int issued; /* latest, from the mds */
186 int implemented; /* implemented superset of issued (for revocation) */
187 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800188 u32 seq, issue_seq, mseq;
189 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700190 unsigned long last_used;
191 struct list_head caps_item;
192};
193
194#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
195#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
196#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
197
198/*
199 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
200 * we first complete any in-process sync writes and writeback any dirty
201 * data before flushing the snapped state (tracked here) back to the MDS.
202 */
203struct ceph_cap_snap {
204 atomic_t nref;
205 struct ceph_inode_info *ci;
206 struct list_head ci_item, flushing_item;
207
208 u64 follows, flush_tid;
209 int issued, dirty;
210 struct ceph_snap_context *context;
211
212 mode_t mode;
213 uid_t uid;
214 gid_t gid;
215
216 void *xattr_blob;
217 int xattr_len;
218 u64 xattr_version;
219
220 u64 size;
221 struct timespec mtime, atime, ctime;
222 u64 time_warp_seq;
223 int writing; /* a sync write is still in progress */
224 int dirty_pages; /* dirty pages awaiting writeback */
225};
226
227static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
228{
229 if (atomic_dec_and_test(&capsnap->nref))
230 kfree(capsnap);
231}
232
233/*
234 * The frag tree describes how a directory is fragmented, potentially across
235 * multiple metadata servers. It is also used to indicate points where
236 * metadata authority is delegated, and whether/where metadata is replicated.
237 *
238 * A _leaf_ frag will be present in the i_fragtree IFF there is
239 * delegation info. That is, if mds >= 0 || ndist > 0.
240 */
241#define CEPH_MAX_DIRFRAG_REP 4
242
243struct ceph_inode_frag {
244 struct rb_node node;
245
246 /* fragtree state */
247 u32 frag;
248 int split_by; /* i.e. 2^(split_by) children */
249
250 /* delegation and replication info */
251 int mds; /* -1 if same authority as parent */
252 int ndist; /* >0 if replicated */
253 int dist[CEPH_MAX_DIRFRAG_REP];
254};
255
256/*
257 * We cache inode xattrs as an encoded blob until they are first used,
258 * at which point we parse them into an rbtree.
259 */
260struct ceph_inode_xattr {
261 struct rb_node node;
262
263 const char *name;
264 int name_len;
265 const char *val;
266 int val_len;
267 int dirty;
268
269 int should_free_name;
270 int should_free_val;
271};
272
273struct ceph_inode_xattrs_info {
274 /*
275 * (still encoded) xattr blob. we avoid the overhead of parsing
276 * this until someone actually calls getxattr, etc.
277 *
278 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
279 * NULL means we don't know.
280 */
281 struct ceph_buffer *blob, *prealloc_blob;
282
283 struct rb_root index;
284 bool dirty;
285 int count;
286 int names_size;
287 int vals_size;
288 u64 version, index_version;
289};
290
291/*
292 * Ceph inode.
293 */
294#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
295#define CEPH_I_NODELAY 4 /* do not delay cap release */
296#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
Sage Weile9964c12010-03-01 15:16:56 -0800297#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
Sage Weilde576062009-10-06 11:31:07 -0700298
299struct ceph_inode_info {
300 struct ceph_vino i_vino; /* ceph ino + snap */
301
302 u64 i_version;
303 u32 i_time_warp_seq;
304
305 unsigned i_ceph_flags;
306 unsigned long i_release_count;
307
308 struct ceph_file_layout i_layout;
309 char *i_symlink;
310
311 /* for dirs */
312 struct timespec i_rctime;
313 u64 i_rbytes, i_rfiles, i_rsubdirs;
314 u64 i_files, i_subdirs;
315 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
316
317 struct rb_root i_fragtree;
318 struct mutex i_fragtree_mutex;
319
320 struct ceph_inode_xattrs_info i_xattrs;
321
322 /* capabilities. protected _both_ by i_lock and cap->session's
323 * s_mutex. */
324 struct rb_root i_caps; /* cap list */
325 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
326 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
327 struct list_head i_dirty_item, i_flushing_item;
328 u64 i_cap_flush_seq;
329 /* we need to track cap writeback on a per-cap-bit basis, to allow
330 * overlapping, pipelined cap flushes to the mds. we can probably
331 * reduce the tid to 8 bits if we're concerned about inode size. */
332 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
333 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
334 unsigned long i_hold_caps_min; /* jiffies */
335 unsigned long i_hold_caps_max; /* jiffies */
336 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
337 int i_cap_exporting_mds; /* to handle cap migration between */
338 unsigned i_cap_exporting_mseq; /* mds's. */
339 unsigned i_cap_exporting_issued;
340 struct ceph_cap_reservation i_cap_migration_resv;
341 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
342 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */
343 unsigned i_snap_caps; /* cap bits for snapped files */
344
345 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
346
347 u32 i_truncate_seq; /* last truncate to smaller size */
348 u64 i_truncate_size; /* and the size we last truncated down to */
349 int i_truncate_pending; /* still need to call vmtruncate */
350
351 u64 i_max_size; /* max file size authorized by mds */
352 u64 i_reported_size; /* (max_)size reported to or requested of mds */
353 u64 i_wanted_max_size; /* offset we'd like to write too */
354 u64 i_requested_max_size; /* max_size we've requested */
355
356 /* held references to caps */
357 int i_pin_ref;
358 int i_rd_ref, i_rdcache_ref, i_wr_ref;
359 int i_wrbuffer_ref, i_wrbuffer_ref_head;
360 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
361 u32 i_rdcache_gen; /* we increment this each time we get
362 FILE_CACHE. If it's non-zero, we
363 _may_ have cached pages. */
364 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
365
366 struct list_head i_unsafe_writes; /* uncommitted sync writes */
367 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
368 spinlock_t i_unsafe_lock;
369
370 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
371 int i_snap_realm_counter; /* snap realm (if caps) */
372 struct list_head i_snap_realm_item;
373 struct list_head i_snap_flush_item;
374
375 struct work_struct i_wb_work; /* writeback work */
376 struct work_struct i_pg_inv_work; /* page invalidation work */
377
378 struct work_struct i_vmtruncate_work;
379
380 struct inode vfs_inode; /* at end */
381};
382
383static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
384{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700385 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700386}
387
388static inline void ceph_i_clear(struct inode *inode, unsigned mask)
389{
390 struct ceph_inode_info *ci = ceph_inode(inode);
391
392 spin_lock(&inode->i_lock);
393 ci->i_ceph_flags &= ~mask;
394 spin_unlock(&inode->i_lock);
395}
396
397static inline void ceph_i_set(struct inode *inode, unsigned mask)
398{
399 struct ceph_inode_info *ci = ceph_inode(inode);
400
401 spin_lock(&inode->i_lock);
402 ci->i_ceph_flags |= mask;
403 spin_unlock(&inode->i_lock);
404}
405
406static inline bool ceph_i_test(struct inode *inode, unsigned mask)
407{
408 struct ceph_inode_info *ci = ceph_inode(inode);
409 bool r;
410
411 smp_mb();
412 r = (ci->i_ceph_flags & mask) == mask;
413 return r;
414}
415
416
417/* find a specific frag @f */
418extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
419 u32 f);
420
421/*
422 * choose fragment for value @v. copy frag content to pfrag, if leaf
423 * exists
424 */
425extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
426 struct ceph_inode_frag *pfrag,
427 int *found);
428
429/*
430 * Ceph dentry state
431 */
432struct ceph_dentry_info {
433 struct ceph_mds_session *lease_session;
434 u32 lease_gen, lease_shared_gen;
435 u32 lease_seq;
436 unsigned long lease_renew_after, lease_renew_from;
437 struct list_head lru;
438 struct dentry *dentry;
439 u64 time;
440 u64 offset;
441};
442
443static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
444{
445 return (struct ceph_dentry_info *)dentry->d_fsdata;
446}
447
448static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
449{
450 return ((loff_t)frag << 32) | (loff_t)off;
451}
452
453/*
454 * ino_t is <64 bits on many architectures, blech.
455 *
456 * don't include snap in ino hash, at least for now.
457 */
458static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
459{
460 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
461#if BITS_PER_LONG == 32
462 ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
463 if (!ino)
464 ino = 1;
465#endif
466 return ino;
467}
468
469static inline int ceph_set_ino_cb(struct inode *inode, void *data)
470{
471 ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
472 inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
473 return 0;
474}
475
476static inline struct ceph_vino ceph_vino(struct inode *inode)
477{
478 return ceph_inode(inode)->i_vino;
479}
480
481/* for printf-style formatting */
482#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
483
484static inline u64 ceph_ino(struct inode *inode)
485{
486 return ceph_inode(inode)->i_vino.ino;
487}
488static inline u64 ceph_snap(struct inode *inode)
489{
490 return ceph_inode(inode)->i_vino.snap;
491}
492
493static inline int ceph_ino_compare(struct inode *inode, void *data)
494{
495 struct ceph_vino *pvino = (struct ceph_vino *)data;
496 struct ceph_inode_info *ci = ceph_inode(inode);
497 return ci->i_vino.ino == pvino->ino &&
498 ci->i_vino.snap == pvino->snap;
499}
500
501static inline struct inode *ceph_find_inode(struct super_block *sb,
502 struct ceph_vino vino)
503{
504 ino_t t = ceph_vino_to_ino(vino);
505 return ilookup5(sb, t, ceph_ino_compare, &vino);
506}
507
508
509/*
510 * caps helpers
511 */
512static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
513{
514 return !RB_EMPTY_ROOT(&ci->i_caps);
515}
516
517extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
518extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
519extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
520 struct ceph_cap *cap);
521
522static inline int ceph_caps_issued(struct ceph_inode_info *ci)
523{
524 int issued;
525 spin_lock(&ci->vfs_inode.i_lock);
526 issued = __ceph_caps_issued(ci, NULL);
527 spin_unlock(&ci->vfs_inode.i_lock);
528 return issued;
529}
530
531static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
532 int touch)
533{
534 int r;
535 spin_lock(&ci->vfs_inode.i_lock);
536 r = __ceph_caps_issued_mask(ci, mask, touch);
537 spin_unlock(&ci->vfs_inode.i_lock);
538 return r;
539}
540
541static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
542{
543 return ci->i_dirty_caps | ci->i_flushing_caps;
544}
Sage Weilafcdaea2009-10-14 14:27:38 -0700545extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700546
547extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
548extern int __ceph_caps_used(struct ceph_inode_info *ci);
549
550extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
551
552/*
553 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
554 */
555static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
556{
557 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
558 if (w & CEPH_CAP_FILE_BUFFER)
559 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
560 return w;
561}
562
563/* what the mds thinks we want */
564extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
565
566extern void ceph_caps_init(void);
567extern void ceph_caps_finalize(void);
Sage Weil85ccce42010-02-17 10:02:43 -0800568extern void ceph_adjust_min_caps(int delta);
Sage Weilde576062009-10-06 11:31:07 -0700569extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need);
570extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx);
571extern void ceph_reservation_status(struct ceph_client *client,
572 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800573 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700574
575static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
576{
577 return (struct ceph_client *)inode->i_sb->s_fs_info;
578}
579
580static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
581{
582 return (struct ceph_client *)sb->s_fs_info;
583}
584
Sage Weilde576062009-10-06 11:31:07 -0700585
586/*
587 * we keep buffered readdir results attached to file->private_data
588 */
589struct ceph_file_info {
590 int fmode; /* initialized on open */
591
592 /* readdir: position within the dir */
593 u32 frag;
594 struct ceph_mds_request *last_readdir;
595 int at_end;
596
597 /* readdir: position within a frag */
598 unsigned offset; /* offset of last chunk, adjusted for . and .. */
599 u64 next_offset; /* offset of next chunk (last_name's + 1) */
600 char *last_name; /* last entry in previous chunk */
601 struct dentry *dentry; /* next dentry (for dcache readdir) */
602 unsigned long dir_release_count;
603
604 /* used for -o dirstat read() on directory thing */
605 char *dir_info;
606 int dir_info_len;
607};
608
609
610
611/*
612 * snapshots
613 */
614
615/*
616 * A "snap context" is the set of existing snapshots when we
617 * write data. It is used by the OSD to guide its COW behavior.
618 *
619 * The ceph_snap_context is refcounted, and attached to each dirty
620 * page, indicating which context the dirty data belonged when it was
621 * dirtied.
622 */
623struct ceph_snap_context {
624 atomic_t nref;
625 u64 seq;
626 int num_snaps;
627 u64 snaps[];
628};
629
630static inline struct ceph_snap_context *
631ceph_get_snap_context(struct ceph_snap_context *sc)
632{
633 /*
634 printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
635 atomic_read(&sc->nref)+1);
636 */
637 if (sc)
638 atomic_inc(&sc->nref);
639 return sc;
640}
641
642static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
643{
644 if (!sc)
645 return;
646 /*
647 printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
648 atomic_read(&sc->nref)-1);
649 */
650 if (atomic_dec_and_test(&sc->nref)) {
651 /*printk(" deleting snap_context %p\n", sc);*/
652 kfree(sc);
653 }
654}
655
656/*
657 * A "snap realm" describes a subset of the file hierarchy sharing
658 * the same set of snapshots that apply to it. The realms themselves
659 * are organized into a hierarchy, such that children inherit (some of)
660 * the snapshots of their parents.
661 *
662 * All inodes within the realm that have capabilities are linked into a
663 * per-realm list.
664 */
665struct ceph_snap_realm {
666 u64 ino;
667 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800668 struct rb_node node;
669
Sage Weilde576062009-10-06 11:31:07 -0700670 u64 created, seq;
671 u64 parent_ino;
672 u64 parent_since; /* snapid when our current parent became so */
673
674 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
675 int num_prior_parent_snaps; /* had prior to parent_since */
676 u64 *snaps; /* snaps specific to this realm */
677 int num_snaps;
678
679 struct ceph_snap_realm *parent;
680 struct list_head children; /* list of child realms */
681 struct list_head child_item;
682
683 struct list_head empty_item; /* if i have ref==0 */
684
685 /* the current set of snaps for this realm */
686 struct ceph_snap_context *cached_context;
687
688 struct list_head inodes_with_caps;
689 spinlock_t inodes_with_caps_lock;
690};
691
692
693
694/*
695 * calculate the number of pages a given length and offset map onto,
696 * if we align the data.
697 */
698static inline int calc_pages_for(u64 off, u64 len)
699{
700 return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
701 (off >> PAGE_CACHE_SHIFT);
702}
703
704
705
706/* snap.c */
707struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
708 u64 ino);
709extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
710 struct ceph_snap_realm *realm);
711extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
712 struct ceph_snap_realm *realm);
713extern int ceph_update_snap_trace(struct ceph_mds_client *m,
714 void *p, void *e, bool deletion);
715extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800716 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700717 struct ceph_msg *msg);
718extern void ceph_queue_cap_snap(struct ceph_inode_info *ci,
719 struct ceph_snap_context *snapc);
720extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
721 struct ceph_cap_snap *capsnap);
722extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
723
724/*
725 * a cap_snap is "pending" if it is still awaiting an in-progress
726 * sync write (that may/may not still update size, mtime, etc.).
727 */
728static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
729{
730 return !list_empty(&ci->i_cap_snaps) &&
731 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
732 ci_item)->writing;
733}
734
735
736/* super.c */
737extern struct kmem_cache *ceph_inode_cachep;
738extern struct kmem_cache *ceph_cap_cachep;
739extern struct kmem_cache *ceph_dentry_cachep;
740extern struct kmem_cache *ceph_file_cachep;
741
742extern const char *ceph_msg_type_name(int type);
Sage Weil07433042009-11-18 16:50:41 -0800743extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
Sage Weilde576062009-10-06 11:31:07 -0700744
745#define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \
746 "%02x%02x%02x%02x%02x%02x"
747#define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \
748 (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7], \
749 (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11], \
750 (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15]
751
752/* inode.c */
753extern const struct inode_operations ceph_file_iops;
754
755extern struct inode *ceph_alloc_inode(struct super_block *sb);
756extern void ceph_destroy_inode(struct inode *inode);
757
758extern struct inode *ceph_get_inode(struct super_block *sb,
759 struct ceph_vino vino);
760extern struct inode *ceph_get_snapdir(struct inode *parent);
761extern int ceph_fill_file_size(struct inode *inode, int issued,
762 u32 truncate_seq, u64 truncate_size, u64 size);
763extern void ceph_fill_file_time(struct inode *inode, int issued,
764 u64 time_warp_seq, struct timespec *ctime,
765 struct timespec *mtime, struct timespec *atime);
766extern int ceph_fill_trace(struct super_block *sb,
767 struct ceph_mds_request *req,
768 struct ceph_mds_session *session);
769extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
770 struct ceph_mds_session *session);
771
772extern int ceph_inode_holds_cap(struct inode *inode, int mask);
773
774extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700775extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800776extern void ceph_queue_vmtruncate(struct inode *inode);
777
778extern void ceph_queue_invalidate(struct inode *inode);
779extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700780
781extern int ceph_do_getattr(struct inode *inode, int mask);
782extern int ceph_permission(struct inode *inode, int mask);
783extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
784extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
785 struct kstat *stat);
786
787/* xattr.c */
788extern int ceph_setxattr(struct dentry *, const char *, const void *,
789 size_t, int);
790extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
791extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
792extern int ceph_removexattr(struct dentry *, const char *);
793extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
794extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
795
796/* caps.c */
797extern const char *ceph_cap_string(int c);
798extern void ceph_handle_caps(struct ceph_mds_session *session,
799 struct ceph_msg *msg);
800extern int ceph_add_cap(struct inode *inode,
801 struct ceph_mds_session *session, u64 cap_id,
802 int fmode, unsigned issued, unsigned wanted,
803 unsigned cap, unsigned seq, u64 realmino, int flags,
804 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800805extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700806static inline void ceph_remove_cap(struct ceph_cap *cap)
807{
808 struct inode *inode = &cap->ci->vfs_inode;
809 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800810 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700811 spin_unlock(&inode->i_lock);
812}
Sage Weil7c1332b2010-02-16 11:39:45 -0800813extern void ceph_put_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700814
815extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100816extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Sage Weilde576062009-10-06 11:31:07 -0700817extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync);
818extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
819 struct ceph_mds_session *session);
820extern int ceph_get_cap_mds(struct inode *inode);
821extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
822extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
823extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
824 struct ceph_snap_context *snapc);
825extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
826 struct ceph_mds_session **psession);
827extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
828 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700829extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
830extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700831
832extern int ceph_encode_inode_release(void **p, struct inode *inode,
833 int mds, int drop, int unless, int force);
834extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
835 int mds, int drop, int unless);
836
837extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
838 int *got, loff_t endoff);
839
840/* for counting open files by mode */
841static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
842{
843 ci->i_nr_by_mode[mode]++;
844}
845extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
846
847/* addr.c */
848extern const struct address_space_operations ceph_aops;
849extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
850
851/* file.c */
852extern const struct file_operations ceph_file_fops;
853extern const struct address_space_operations ceph_aops;
854extern int ceph_open(struct inode *inode, struct file *file);
855extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
856 struct nameidata *nd, int mode,
857 int locked_dir);
858extern int ceph_release(struct inode *inode, struct file *filp);
859extern void ceph_release_page_vector(struct page **pages, int num_pages);
860
861/* dir.c */
862extern const struct file_operations ceph_dir_fops;
863extern const struct inode_operations ceph_dir_iops;
864extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
865 ceph_snapdir_dentry_ops;
866
867extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
868extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
869 struct dentry *dentry, int err);
870
871extern void ceph_dentry_lru_add(struct dentry *dn);
872extern void ceph_dentry_lru_touch(struct dentry *dn);
873extern void ceph_dentry_lru_del(struct dentry *dn);
874
875/*
876 * our d_ops vary depending on whether the inode is live,
877 * snapshotted (read-only), or a virtual ".snap" directory.
878 */
879int ceph_init_dentry(struct dentry *dentry);
880
881
882/* ioctl.c */
883extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
884
885/* export.c */
886extern const struct export_operations ceph_export_ops;
887
888/* debugfs.c */
889extern int ceph_debugfs_init(void);
890extern void ceph_debugfs_cleanup(void);
891extern int ceph_debugfs_client_init(struct ceph_client *client);
892extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
893
894static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
895{
896 if (dentry && dentry->d_parent)
897 return dentry->d_parent->d_inode;
898
899 return NULL;
900}
901
902#endif /* _FS_CEPH_SUPER_H */